Reward models (RMs) are the moral compass of LLMs – but no one has x-rayed them at scale. We just ran the first exhaustive analysis of 10 leading RMs, and the results were...eye-opening. Wild disagreement, base-model imprint, identity-term bias, mere-exposure quirks & more: 🧵 https://t.co/AFGGNVuHNi

Decoding the Black Box of Compensation, Part II Introducing https://t.co/OYYpRGHGmz, a tool to algorithmically estimate fair compensation — for any title, at any stage of company, anywhere in the world. After thousands of regressions, analyses and backtests, we’ve determined that there are... See more

Starting to see the first serious economic analysis attempts to grapple with what AGI might mean. I appreciate that this piece embraces scenarios, we don’t know if or when AGI might happen. But wow that wages graph is something else. https://t.co/0qc3onJZM3